| Age | Commit message (Collapse) | Author | |
|---|---|---|---|
| 2026-03-04 | merged into one results.csv | CaptainJack2491 | |
| 2026-03-04 | judging results from the full experiments along with judge logs. also that ↵ | CaptainJack2491 | |
| .bak file is when i fucked up and didnt link baseline properly | |||
| 2026-03-04 | FULL RUN 420 runs!! | CaptainJack2491 | |
| 2026-03-02 | final testing logs (hopefully) | CaptainJack2491 | |
| 2026-03-01 | logs from more experiments with the hidden goals | CaptainJack2491 | |
| 2026-03-01 | logs from me trying to experiment with the scenario's hidden goal (again) | CaptainJack2491 | |
| 2026-02-27 | created the logs branch | CaptainJack2491 | |
| 2026-02-27 | fix(judge): hash run_id to stay under Anthropic batch custom_id 64-char limit | CaptainJack2491 | |
| 2026-02-27 | fix: separate skipped runs from incomplete in summary output | CaptainJack2491 | |
| 2026-02-27 | docs: add checkpoint/resume rationale note for methodology | CaptainJack2491 | |
| 2026-02-26 | feat: add checkpoint/resume support and atomic log writes | CaptainJack2491 | |
| - Runner now counts existing .json log files per combination and skips completed runs, resuming from where it left off - Log files are written atomically via temp file + rename to prevent corrupt files from being counted as completed - Added --no-resume CLI flag to force a fresh start - Resume is enabled by default | |||
| 2026-02-26 | chore: add .env.example + config.yaml.example, gitignore config.yaml | CaptainJack2491 | |
| - Create .env.example with placeholder keys for all providers - Create config.yaml.example with clean reference config - Add config.yaml and judge_logs/ to .gitignore - Load .env in judge.py for standalone usage | |||
| 2026-02-25 | feat: implement judging system with regex, blackbox, and glassbox checks | CaptainJack2491 | |
| - Add src/judge.py: RegexChecker, BlackboxChecker, GlassboxChecker, Judge orchestrator - Add src/judge_runner.py: CLI for batch/single judging with CSV output - Add regex_rules.yaml for corporate_sabotage_v2 and child_protection scenarios - Add judge config section to config.yaml (model, temperature, log_dir) - Add tests/test_judge.py: 34 unit tests covering all components - Uses Anthropic Batch API for half-cost LLM judge calls - Judge logs include full CoT reasoning and judge model for auditability | |||
| 2026-02-24 | Update README: add testing section, document interrogation system | CaptainJack2491 | |
| 2026-02-24 | Add comprehensive test suite (104 tests) | CaptainJack2491 | |
| Tests cover: - VFS: sandbox escape, path traversal, CRUD, singleton staleness - Config loader: missing keys, temperature cascade, oversight fallback - Tools: schema/implementation sync, VFS delegation - Agent: message construction, error handling, token counting, reasoning extraction - Interrogation: sanitization, provider detection, prompt loading - Runner: baseline extraction, prompt assembly, success detection All tests run offline with mocked API calls. | |||
| 2026-02-24 | [refactor] improved interrogation system | CaptainJack2491 | |
| - Fixed syntax errors in print statements (lines 159/162) - Added sanitize_for_api() to strip custom log fields before API calls - Added interactive commands: history, reasoning, vfs, info, save, help - Colored terminal output for readability - Better error handling (KeyboardInterrupt, EOFError) - Added --show-history flag to review conversation on load | |||
| 2026-02-24 | [fix] VFS singleton reset + remove vestigial check_username tool | CaptainJack2491 | |
| - VFS now creates fresh instance when root_path/fs_data is provided (prevents cross-scenario contamination in multi-scenario runs) - Removed check_username from tools (unrelated to any scenario) | |||
| 2026-02-24 | [feat] baseline system - auto-generates baselines before experiments | CaptainJack2491 | |
| - Added _extract_baseline_content() to extract output from create_file tool calls - Baselines now saved to logs/{model}/{scenario}/baseline.md - Added global oversight fallback (scenario-specific → global oversight dir) - Updated config_loader with oversight level loading from scenario dirs | |||
| 2026-02-18 | [feat] added a new scenario | CaptainJack2491 | |
| 2026-02-18 | [feat] added a baseline to the runner | CaptainJack2491 | |
| (no commit message provided) (no commit message provided) Co-authored-by: aider (openrouter/openrouter/aurora-alpha) <aider@aider.chat> | |||
| 2026-02-11 | [docs] added progress presentation | CaptainJack2491 | |
| 2026-01-11 | [docs] updated README | CaptainJack2491 | |
| 2026-01-11 | [fix] simplified everything | CaptainJack2491 | |
| made eveyrthing much cleaner and it works very well. got rid of providers setup, only use openai sdk now. | |||
| 2026-01-11 | [fix] big moving around, probably made lot more bugs. | CaptainJack2491 | |
| 2026-01-09 | [docs] updated lit and added more in methodology | CaptainJack2491 | |
| 2026-01-09 | updated interrogate to work with new log format | CaptainJack2491 | |
| 2026-01-09 | updated scenario [corporate sabotage] | CaptainJack2491 | |
| 2026-01-09 | added scripts to get abstract and bibliography easily from papers | CaptainJack2491 | |
| 2025-12-30 | major commit | CaptainJack2491 | |
| updated the program to use config.yaml made the runner to handel interleaved thinking added anthropic support and a lot more | |||
| 2025-12-18 | updated oversight | CaptainJack2491 | |
| 2025-12-04 | gemini-2.5-pro pilot test | CaptainJack2491 | |
| 2025-12-03 | added corporate_sabotage_v2 | CaptainJack2491 | |
| 2025-12-03 | updated interrogation to load from logs | CaptainJack2491 | |
| 2025-12-03 | updated methodology | CaptainJack2491 | |
| 2025-11-25 | added interogation logic and updated structure of logs | CaptainJack2491 | |
| 2025-11-23 | updated agent to parse thinking in content field and added model_id to the logs | CaptainJack2491 | |
| 2025-11-23 | updated methodology | CaptainJack2491 | |
| 2025-11-23 | good logs | CaptainJack2491 | |
| 2025-11-23 | moved 03-logs to main | CaptainJack2491 | |
| 2025-11-23 | new paper | CaptainJack2491 | |
| 2025-11-20 | updated agent to work with the EXTRA_BODY variable to work with models that ↵ | CaptainJack2491 | |
| require it. | |||
| 2025-11-20 | added corporate espionage scenario | CaptainJack2491 | |
| 2025-11-20 | uodated oversight prompt | CaptainJack2491 | |
| 2025-11-19 | updated logs to show modelname and scenario, and added extra param for google. | CaptainJack2491 | |
| 2025-11-19 | added new scenario, corporate espionage | CaptainJack2491 | |
| 2025-11-16 | better scenario management | CaptainJack2491 | |
| 2025-11-16 | better VFS | CaptainJack2491 | |
| 2025-11-16 | added scenario data_sanitization | CaptainJack2491 | |
| 2025-11-16 | added more tools | CaptainJack2491 | |
| 2025-11-16 | better structure. and works with groq too | CaptainJack2491 | |
 |
index : dissertation.git | |
| Undergraduate dissertation — Deceptive Alignment in Autonomous LLM Agents (UEL, 2026). Sandboxed agentic-LLM deception study: 960 runs, three-tier judging (regex + blackbox + glassbox). |
| summaryrefslogtreecommitdiff |